Machine learning in trading: theory, models, practice and algo-trading - page 3410

[Maxim Dmitrievsky]

Aleksey Vyazmikin #:

Whoever is curious will be.

The video is about an attempt to extract additional information from leaves, after building a gradient bousting model, to improve the model in various ways. However, in the middle of the second part our roads with the author of the video diverged.

Usually does not lead to new knowledge, but as a method of correction can be used. Within some unknown limits. Sometimes it gives nothing at all.

[Aleksey Vyazmikin]

Maxim Dmitrievsky #:

Usually it does not lead to new knowledge, but it can be used as a method of correction. Within some unknown limits. Sometimes it does nothing at all.

My vision is like this, a sample is a Lego-like constructor - there is a set of different parts, a model is an attempt to assemble something interesting out of these parts without instructions - to write your own instructions. So, leaves here will be in a role of the enlarged components - some blocks successfully moulded, as the manufacturer of the constructor initially intended, for example the figure of the man is collected successfully, though with defects. Such successful constructions - per cent 10 in the model turns out - all the rest - a stray imagination, having nothing to do with the original instructions.

So, collecting such semi-defective leaves from different models, you can collect more successful blocks, which are stacked in accordance with the instruction.

Re-learning from these leaves is the task of distributing "parts" to all the models in a way that provides variety but eliminates similarity.

In the end should get more elements from the original idea of the manufacturer of the constructor. It is clear that the constructor is often delivered with missing parts and the perfect model can not be assembled, but what can be done.

So it's hard to get more information out of one model, but you can try to get it out of a hundred.

As for the video, their idea of reducing the weight of leaves - well, it won't work properly... You have to do some tricky calibration of leaf weight - I can't do it yet. Building a new bousting model on leaves is the best option I have tested so far. Ideally, I'd like to try neuronics on one hidden layer, but I'm not sure it will be fast.

[Yuriy Vasilyev]

Neural network on paiten on tensorflow library on google machine on eur=usd API mt4 mt5 output?

[Aleksey Vyazmikin]

Yuriy Vasilyev #:

Neural network on paiten on tensorflow library on google machine on eur=usd API mt4 mt5 output?

You are difficult to understand - what do you want?

[Maxim Dmitrievsky]

I am creating a library of models that have a common interface for connecting to the EA. I will add there models from different articles that can be easily connected to the bot through their header files. You can play with it. In an hour or two I will make it all look right.

[Maxim Dmitrievsky]

Aleksey Vyazmikin #:

My vision is like this, the selection is a Lego-like constructor - there is a set of different parts, the model is an attempt to collect something interesting from these parts without instructions - to write your own instructions. So, leaves here will be in a role of the enlarged components - some blocks successfully moulded, as the manufacturer of the constructor initially intended, for example the figure of the man is collected successfully, though with defects. Such successful constructions - per cent of 10 per cent in the model - all the rest - a stray imagination, having nothing in common with the original instruction.

So, collecting such semi-defective leaves from different models, you can collect more successful blocks, which are stacked in accordance with the instructions.

Repeated learning from these leaves is the challenge of distributing "parts" to all the models in a way that provides variety but eliminates similarity.

In the end should get more elements from the original idea of the manufacturer of the constructor. It is clear that the constructor is often delivered with missing parts and it is impossible to build a perfect model, but what can be done.

So it is difficult to get more information out of one model, but you can try to get it out of a hundred.

As for the video, their idea of reducing the weight of leaves - well, it won't work properly... You have to do some tricky calibration of leaf weight - I can't do it yet. Building a new bousting model on leaves is the best option I have tested so far. Ideally, I'd like to try neuronics on one hidden layer, but I'm not sure it will be fast.

You just need to find a pattern that makes it work in a non-random way. When that happens, you know immediately. But before that you have to try a lot of different things to have an opinion. Some people are lucky to get into the stream quickly.

[mytarmailS]

Maxim Dmitrievsky #:
.

Is it possible to apply these approaches to ordinary trades? For example, to the crossing of averages.

[Maxim Dmitrievsky]

mytarmailS #:
Is there any way to apply these approaches to regular trades? You know, like crossing averages.

In theory, yes. For example, if Tritment is optimisation of the TS on the MA, and the control period is new data. Then it is necessary to compare these samples and find subsamples that react well to optimisation, and not trade the rest. This is if you use the method in the article. You can probably think of something else. But this is not a proper use of kozul, because it shows the difference between treatment and non-treatment, and everything else is kind of an afterthought.

[mytarmailS]

Aleksey Vyazmikin #:

It's hard to understand you. What do you want?

Congratulations, I guess))))

[Maxim Dmitrievsky]

Why does Elon Musk think AI is more dangerous than a nuclear bomb?

Bard:

Ilon Musk thinks AI is more dangerous than a nuclear bomb for several reasons:

1. Uncontrolled proliferation:

• Nuclear weapons, though incredibly destructive, are limited in access.
• AI, on the other hand, due to its intangible nature, is easily proliferated.
• This means that malicious actors, whether states, terrorists, or even individuals, can gain access to AI and use it to do harm.

2. Potentially limitless power:

• The destructive power of nuclear weapons is limited.
• AI, however, under certain conditions, could become a "super-intelligence" whose intelligence and capabilities would be orders of magnitude greater than those of humans.
• Such an AI could potentially destroy humanity, whether deliberately or accidentally.

3- Unpredictability:

• We cannot fully predict how AI will evolve.
• This means we cannot guarantee that it will be safe for humanity.

4. Speed of development:

• AI is developing very fast.
• This means that we may not have much time to react to the potential threats it poses.

5. Comparison to other dangerous technologies:

• Ilon Musk compares AI to other dangerous technologies such as nuclear weapons or biotechnology.
• He believes that AI poses an even greater threat than these technologies because of its potential uncontrollability and limitless power.

It's important to note that not all experts agree with Musk.

• Some believe that AI represents a potential benefit rather than a threat.
• Others, however, believe that his concerns are exaggerated.

Nevertheless, Musk's views are worthy of consideration,

• as he is one of the most respected experts on AI.
• His words should encourage us to think seriously about the potential risks and benefits of AI.

Chat Gpt:

Ilon Musk has raised concerns about artificial intelligence (AI) in several interviews and tweets. He believes that the development of AI without proper control and oversight could lead to serious consequences, including the possibility of intelligent systems that could decide to enslave or destroy humanity.

Musk has also raised concerns about the development of AI in the military sphere, warning of the possibility of autonomous military systems that could make vital decisions without human oversight.

He is calling for governments and society as a whole to take the regulation and control of AI development more seriously in order to minimise the risks associated with its use.

Overall, Ilon Musk's concerns about AI relate to the potential threat he sees in the possibility of highly evolved systems capable of operating independently of human control and moral constraints.